What challenges did Cisco face before AI implementation?

Webex LLMs had grown to hundreds of gigabytes and were difficult to start and scale when embedded in application containers. Cisco needed better resource utilization and faster deployment for contact-center transcript analysis and customer support workflows.

What was the impact of the AI solution?

The platform can handle hundreds of inferences per minute at peak and overnight batch jobs reliably. Cisco says the collaboration helped reduce foundation model deployment costs by 50% on average and latency by 20% on average.

Source

Evidence: Low35/100

Cisco Webex Contact Center Topic Analytics: migrating LLM workloads to SageMaker Inference

Use case typeCustomer experience analyticsUpdated Jun 13, 2026

Cisco's Webex AI team migrated LLM workloads from application containers to Amazon SageMaker Inference to improve speed, scalability, and price-performance. The solution powers Webex Contact Center topic analytics and other AI experiences, including transcript analysis, conversational assistants, and meeting summaries.

Organization: Cisco
Industry: Tech & Comms
Location: United States
Published: August 2024

Reported outcomes

Cost: −50%

Cost savings

Cost: −20%

Catalog median for cost savings deployments: −40% across 171 reported metrics. Compare benchmarks →

Why do we believe this?Outcome claims, sources, and evidence checks

Normalized claim

Cost: 50% decrease

AWS BlogAug 8, 2024Blog postInferred claimLow evidence strength

Cisco says the collaboration helped reduce foundation model deployment costs by 50% on average and latency by 20% on average.

Normalized claim

Cost: 20% decrease

AWS BlogAug 8, 2024Blog postInferred claimLow evidence strength

Cisco says the collaboration helped reduce foundation model deployment costs by 50% on average and latency by 20% on average.

Why do we believe this deployment?Customer identity, provider attribution, maturity, and source checks

Customer: Cisco, Webex, Webex Contact Center
Provider: AWS
Maturity: Unknown
Linked source: AWS Blog

No explicit deployment-stage evidence found.

Customer identity supportedSource describes one deploymentMaturity evidence evaluated

Primary read

Use case focus

Showing 3 of 4

1Contact Center Analytics
2Conversational AI
3Model Inference Optimization

Cisco decoupled model hosting from applications and moved LLM inference to Amazon SageMaker Inference.
The architecture uses an LLM proxy on Amazon EKS, auto scaling, and a three-model pipeline for topic analytics with FLAN-T5 call-driver extraction, clustering, and topic labeling.
The stack also uses Amazon Bedrock as an optional model provider and AWS infrastructure services for ingress and observability.

Technologies

Amazon SageMaker Inference Amazon Bedrock Amazon EKS Application Load Balancer Amazon Managed Service for Prometheus Amazon RDS FLAN-T5 Llama2-13b-chat NVIDIA Triton Inference Server

The platform can handle hundreds of inferences per minute at peak and overnight batch jobs reliably.
Cisco says the collaboration helped reduce foundation model deployment costs by 50% on average and latency by 20% on average.

Architecture

Cisco's WxAI architecture decouples model hosting from Webex applications by moving LLMs to Amazon SageMaker Inference. An LLM proxy runs on Amazon EKS and routes requests to multiple models. Webex Contact Center Topic Analytics uses a three-model pipeline: FLAN-T5 for call-driver extraction, clustering over embeddings, and a generative topic-labeling model. The architecture uses SageMaker auto scaling, Amazon Bedrock as an optional provider, and AWS services for ingress, logging, metrics, and data storage.

Sources & evidence1

Evidence: Low35/100Evidence strength